AI Weekly Malaysia

Meta released Muse Glimmer, a 30B parameter open-weights model under a clean Apache 2.0 license, optimized for agentic task completion, tool use, and multi-step reasoning. Simon Willison tested it locally via LM Studio (18.16 GB quantized), ran it as a coding agent against a Datasette checkout, and confirmed it works as a vision model for image description.

Why: If you want a locally-runnable model for agentic coding and tool-use workflows, Muse Glimmer's Apache 2.0 license removes the Llama licensing friction for commercial use, and its 30B size means it fits on machines with 32GB+ RAM alongside other applications. Test it with your own coding-agent scaffolding before committing—Willison needed a patch for LLM 0.32 compatibility, so expect integration rough edges.

LiquidAI released LFM2.5-VL-3B, a 3.1B parameter vision-language model designed for on-device/edge use, pairing a SigLIP2 400M vision encoder with their LFM2.5-2.6B text backbone. It was pre-trained on ~34T tokens with 4x more vision data than prior versions, supports 128K vocabulary for non-Latin scripts, and adds screen/UI understanding, object grounding, multi-image input, and function calling. Benchmarks show it leading its size class on real-world image tasks (RealWorldQA 73.1, MMStar 63.3) against comparably-sized models from Qwen, InternVL, and Gemma.

Why: If you are building on-device apps that need document/screen understanding or vision-grounded function calling without cloud API latency or cost, this is a concrete 3B model worth benchmarking against Qwen3.5-2B or InternVL 3.5 2B for your use case. The function-calling capability in vision-text contexts is the differentiator to test, since most small VLMs struggle there.

British Transport Police is expanding its live facial recognition trial to the London Underground, starting at Victoria station and rotating through stations until November 2026. The system uses NEC's NeoFace M40 algorithm to scan faces against a watchlist, generating alerts that officers must review before acting. Privacy campaigners like Big Brother Watch call the expansion dystopian, noting errors have already led to innocent people being detained and that ethnic minorities face higher misidentification risk.

Why: For builders working on computer vision or AI deployment, this is a concrete case study of facial recognition rolling into routine public infrastructure at scale (3.7M daily Tube journeys), with documented misidentification and bias issues. If you ship biometric or surveillance-adjacent tech, the operational and reputational risks shown here—false positives, demographic bias, civil liberties backlash—are the ones you need to design around, not abstract ethics frameworks.

Japan's NEC will test a parking system in September, developed with UrbanChain, that uses cameras and AI video analysis to detect when a driver exits their car and only starts charging at that moment—rather than charging from facility entry. NEC is also exploring data services around occupant exit counts and visitor trend analysis, and claims safety benefits from detecting drivers who remain in cars for extended periods.

Why: This is a vendor pilot of a computer-vision parking product with no release date, pricing, or API. For builders, the only concrete signal is NEC's mention of 'data utilization services' for commercial facilities—suggesting future B2B analytics products around vehicle occupancy data—but nothing here requires action or a decision today.